Word

ISSN: 0043-7956 (Print) 2373-5112 (Online) Journal homepage: www.tandfonline.com/journals/rwrd20

Linguistics, Literature, and the Concept of Style

J. M. Ellis

To cite this article: J. M. Ellis (1970) Linguistics, Literature, and the Concept of Style, Word, 26:1,
65-78, DOI: 10.1080/00437956.1970.11435581
To link to this article: https://doi.org/10.1080/00437956.1970.11435581

Published online: 16 Jun 2015.

Submit your article to this journal

Article views: 14890

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rwrd20
~ Af. ELLIS-----------------------------

Linguistics, Literature, and the

Concept of Style

I
In this article I wish to argue that the concept of style, which is and has
long been central in discussion of the relation between linguistic analysis
and literary study, in fact serves only to confuse and obscure that relation-
ship; that it is part of an ordinary language vocabulary entailing an unre-
ftective common-sense analysis of language and its effects which does not
accord with more recent research into language and literature; and that
once the concept, and the theory oflanguage which it implies, is abandoned,
and the effects of language usually referred to as style are dealt with in a
different way, many current problems in the relation of linguistics to
literary study disappear. Rather than continuing to be puzzled by the
elusiveness of the concept and brooding over its "real" meaning or mean-
ings, we should recognize that it has no usefulness in the study of language
and literature.

II
Linguists of all schools have agreed on placing the concept of style at the
center of the linguistic study of literature. In Britain, Halliday equates
stylistics and the linguistic study of literature, 1 and the volume produced
there in 1966 dealing with the relation of linguistics and literature was en-
titled Essays on Style and Language.z In America, the volume Style in
Language 3 has been the most weighty document in the linguistic study of
literature, while writers not represented in that compendium have
also assumed the notions of style and stylistics to be their starting-points. 4
Whether one takes linguists writing in the framework of Chomsky, like
1 M. A. K. Halliday, "Categories of the Theory of Grammar," Word, XVII (1961),
242.
2 Roger Fowler, ed. (London, 1966). Cf. also the volume by Nils Erik Enkvist, John
Spencer, and Michael J. Gregory, Linguistics and Style (London, 1964).
3 Thomas A. Sebeok, ed. (Cambridge, Mass., 1960).
4 Notably Michael Riffaterre, "Criteria for Style Analysis," Word, XV (1959), 154-
174, and "Stylistic Context," Word, XVI (1960), 207-218.
65
66 J. M. ELLIS

Ohmann5 or more conservative structuralists like Hill, 6 the terms style

and stylistics are central.
The common assumption with regard to the starting-point of the inquiry
has not been matched by much else in the way of common ground, and no
theory of style has seemed to be promising to any large body of opinion;
critical comment on most of the contributions to Style in Language, for
example, has not been enthusiastic.? Ohmann, writing in 1964, finds
stylistics "in a state of disorganization," the subject "remarkably unen-
cumbered by theoretical insights," and the concept of style itself, even after
his own attempt to explicate it, "elusive." 8 It would be difficult to disagree
with these judgments, in spite of the amount of work that has recently
been devoted to the subject.
Most stylisticians believe, Riffaterre claimed with some justice in 1959,
that a stylistic device is discovered in its being deviant from the linguistic
norm.9 The limitations of this view are easily exposed by Riffaterre,
Wellek, 10 and others; to take only the most simple and immediate argu-
ment against it, not all deviation is style, and not all style is deviation. In
order to pursue the deviation model further, Levin 11 agrees that it can
only cover a limited area of poetics (one case among three). Even if no
other objection to Levin's theory were available, however, it would already
be a serious limitation of modern research that the most common model
used in the linguistic analysis of literature was only applicable to one case
in three. Riffaterre,t2 after rejecting the deviation model, advances a

s Cf. Richard Ohmann, "Generative Grammars and the Concept of Literary

Style," Word, XX (1964), 423-439.
6 Cf. Archibald A. Hill, "Poetry and Stylistics," address given at the University of
Virginia on September 21, 1956, in the Peter Rushton Seminars in Contemporary
Prose and Poetry, and circulated privately.
7 Roman Jakobson's closing statement (pp. 350-377) is a brilliant essay, but Jakobson
avoids the term stylistics, preferring instead poetics.
s Ohmann, pp. 423, 425, and 439. The same judgment is made by Spencer and
Gregory in their essay "An Approach to the Study of Style" in Linguistics and Style;
style is "recognizable but elusive" (p. 59). R. A. Sayee also finds the term in a state of
confusion in the opening remarks of his "The Definition of the Term 'Style,'" Proceed-
ings of the International Comparative Literature Association, 3rd Congress (1962),
pp. 156-166.
9 Riffaterre, "Criteria," p. 167.
10 See Rem! Wellek's closing statement (pp. 408-419) in Style in Language.
11 Samuel R. Levin, "Deviation-Statistical and Determinate-in Poetic Language,"
Lingua, XII (1963), 276-277. Levin's reply to Wellek's attack on the deviation model
misses his point. Wellek's objection is not that the deviation model leads us beyond
grammar, but that such a treatment is illegitimately atornistic.
12 Riffaterre, "Criteria," p. 169.
 LINGUISTICS, LITERATURE, AND THE CONCEPT OF STYLE 67

view which is not entirely dissimilar to it and which is subject to some of the
same criticisms: style for him is a device which deviates from the norm of the
context. While this is a much improved analysis of the exploitation of
expectations established in a text, it nonetheless cannot deal with those
instances where style strikes us as residing in the use of exactly the right
word for the context, or where it is the product of a homogeneous text. The
style of a text cannot always be discussed in terms of individual devices
which are isolable and segmentable.B
There is a widespread feeling among critics of literature that stylistics
brandishes a formidable apparatus of linguistic terminology, some taken
over from general linguistics, some coined for the occasion, but that little
of value has so far been contributed to the study of literature. The reason
for this state of affairs, I now wish to argue, is to be traced to the concept
used as the basis of the discussion.

Ill
Though the concept of style is receiving more and more complex defini-
tions, we should not forget that its origin is in ordinary, unreflective
language. Its meaning in ordinary language can be found in any dictionary.
Common to all definitions I have ever seen in such works is an explanation
that it refers to a way, manner, or form of doing or saying something.l 4 It
is thus in ordinary usage primarily a dualistic concept which distinguishes
two elements in a piece of language. Ohmann concludes from this that
"the theorist of style is confronted by a kind of task that is commonplace
enough in most fields: the task of explicating and toughening up for rigor-
ous use a notion already familiar to the layman."15 Yet in saying this he
has jumped a step: there is a stage in an investigation where we must decide
either that an ordinary language concept can be tightened up, defined
rigorously and made useful for a precise investigation, or that it is too
misleading and confused to be useful. There is no a priori guarantee what
the answer will be; it depends on the particular concept. An example of an
13 An unusually direct statement that style can be segmented is by Waiter A. Koch,
"On the Principles of Stylistics," Lingua, XII (1963), 416: "Any text whatsoever may
be subjected to a stylistic segmentation."
14 I do not here ignore either the very great efforts which have been devoted to the
definition and redefinition of style or the multiplicity of meaning found in discussions
of it. See, for example, Enkvist's essay "On Defining Style" in Linguistics and Style.
Such discussions have resulted in numbers of competing "definitions" of style and
in numerous laments that "style has proved notoriously hard of stringent definition"
(Enkvist, p. 54). I shall argue below that these efforts have been misguided and the la-
ments, the result of a mistaken notion of definition.
15 Ohmann, p. 423. The same kind of argument is found in Spencer and Gregory, p. 59.
68 J. M. ELLIS

ordinary language term which, like style, had a clear field of reference,
but which has been abandoned by specialists, is madness; this has been
replaced by the term mental illness. The second term has then gradually
invaded ordinary language, and it may even replace the first here too. The
extent of the field of reference is not the issue here: it is the same or very
nearly the same for both terms. The change in terms involves a different
attitude by the speaker, a different classification of the phenomenon con-
cerned, and indeed a different theory of its basis and of ways of dealing
with it. The ordinary language term was part of an ordinary language and
common-sense theory. The theory needed replacing; the term, therefore,
also needed replacing.
This choice that we have wht:n dealing with an ordinary language con-
cept has been largely ignored by both linguists and critics. "A style is a way
of writing-that is what the word means," writes Ohmann, 16 as if no other
issues were involved. Riffaterre, at least, does offer a rigorous definition,
and so wholeheartedly takes the first of the alternatives which I mentioned
above, though again without considering that a choice was available:
"Style is understood as an emphasis (expressive, affective or aesthetic)
added to the information conveyed by the linguistic structure, without
alteration of meaning. Which is to say that language expresses and that
style stresses .... "17 This is certainly a rigorous and uncompromising defini-
tion, but it could not be said to be a violation of the ordinary language use
of "style." Yet, for critics of literature, it would seem like a reductio
ad absurdum of the concept of style. For, as Ohmann notes,
the relevant division between fixed and variable components in literature is
by no means so obvious. What is content, and what is form, or style? The
attack on a dichotomy of form and content has been persistent in modem
criticism; to change so much as a word, the argument runs, is to change the
meaning as well. This austere doctrine has a certain theoretical appeal, given
the supposed impossibility of finding exact synonyms, and the ontological
queerness of disembodied content-propositions, for instance-divorced
from any verbal expression. Yet at the same time this doctrine leads to the
altogether counterintuitive conclusion that there can be no such thing as
style, or that style is simply a part of content. 1B
Here, it seems to me, Ohmann has stated an important reason for the
critics' reluctance to abandon the concept of style, rather than having
advanced a valid argument to support this reluctance. For ifl were to trans-
16 Ohmann, p. 423. Koch's ("The Principles of Stylistics") beginning shows a similar
reliance on the concept of style as an irremovable one.
17 Riffaterre, "Criteria," p. 155.
18 Ohmann, p. 427.
 LINGUISTICS, LITERATURE, AND THE CONCEPT OF STYLE 69
late Ohmann's comment into the context of the analogy I have already
used, I should end up with a statement like this: "Doctors have consist-
ently attacked the dichotomy of madness and illness. To be mad, they say,
is to be sick, and subject to treatment. This austere argument has a certain
theoretical appeal. Yet it leads to the counterintuitive conclusion that there
is no such thing as madness, or that madness is simply part of sickness."
The obvious lack of logical consequence in this adaptation of Ohmann
shows, I hope, how he is methodologically mistaken. It is not unusual for
research to make assumptions which are counterintuitive; intuition is, after
all, the repository of the status quo of knowledge. But, in any case, the
critics' line of argument does not lead to the conclusion which Ohmann
draws from it, any more than it follows from my adapted version that mad-
ness does not exist. The phenomena referred to as madness and style
continue to exist and to exist distinctively; what is questioned in both cases
is the usefulness of the popular concepts concerned. Yet we must here note
a fundamental difference in the two situations. Doctors have rejected the
concept of madness because they rejected the theory which it entailed, but
the critics to whom Ohmann refers have by contrast done something very
strange: they have kept the concept, but rejected the theory which is
inseparable from it. If they have done this, though, it is surely because they
too have made Ohmann's mistake: they have thought that to abolish the
concept would be to abolish the phenomenon. The rather schizoid position in
which these critics find themselves as a result is a point to which I shall
return: I want now to turn to some reasons for the rejection of the theory
inherent in the term style.
Unless it £an be shown that an ordinary language concept can usefully
be adapted for the purposes of a precise investigation, there is, I have
argued, no obligation to keep it as far as the investigator is concerned.
Basic to the concept of style in language is a dualistic model. There are
two possible ways in which the term might be adapted for scholarly use:
either a strict dualism in which the word indicates a split of meaning and
manner or that proposed by Koch in which meaning is split into two
separable kinds of meaning.19
A complication which immediately arises is the uncertainty within
linguistics of the notion of meaning. I want here, however, to assume only
one point, that a crude reference theory of meaning will not work: to use
my analogy once more, madness and mental illness may refer to the same
things, but their meanings are quite different. The difference of meaning
is established by the difference of attitude on the part of the speaker and

19 Koch, p. 412.
70 J. M. ELLIS

by the different classification of the phenomena-and these expressions are

obviously not merely stylistically different.
A set of words which would very clearly involve the ordinary language
notion of style would be start, begin, and commence. These three, we
should normally say, mean the same thing, but differ stylistically. They
belong to different stylistic levels. Nonetheless, it seems to me that the
basis of these stylistic levels is one of meaning. It is true, in a crude way
which suits ordinary discourse, that the "same thing" is involved with all
three-but not exactly the same thing. We separate the usages of these
words so that proceedings commence, races start, and children begin to
play. In one respect the same thing happens. However, in others, different
things happen. A law court is not a horse race. We can refer to a "formal
level oflanguage," only because there are formal things and situations, like
law courts. The use of one of the terms rather than the other means that the
speaker has taken a different attitude to the event concerned and that he
has classified it in a particular way (i.e., what he has said means something
which is distinctively different from what he might have said using another
of the three terms). The use of the word commence in relation to a dance,
for example, may well give enough information about that event to dissuade
devotees of certain types of dancing from attending it. To speak strictly,
therefore, these three words ought not to be called three ways of saying the
same thing: they are three closely related but different words and three
closely related but different ideas. They have an overlap of meaning that is
considerable, but there is an area of meaning that is distinct. Suppose that
we had a word in our language which was neutral with regard to all except
the common area of the three and that that word were inception. Then com-
mence could be defined as 'inception in a formal situation.' The relation
of these two terms would be that of genus to species (i.e., like that of tree
to elm). It is obvious that the relation of oak to elm would still involve
common and different elements of meaning even if we had no word tree;
the relation would not be meaning plus style, however large the common
element. The same is the case with begin, start, and commence, and with
all other such systems of related words.
The difference between very big and enormous might well be termed
one of emphasis; but that is just as much a change of meaning as
the change between big and very big. Where an extra word is used to
differentiate two expressions, we tend in ordinary discourse to say that
meaning has been added, while when an expression is replaced by another,
perhaps more differentiated, one, an entirely parallel change of meaning
tends to be called style. The word formal is one that has a recognizable
meaning; but when its area of meaning is absorbed into another word and
 LINGUISTICS, LITERATURE, AND THE CONCEPT OF STYLE 71

is used to differentiate that word from another which is closely related, the
differentiation is commonly termed stylistic. Yet there seems to be no
theoretical reason for the different descriptions of these two cases. Simi-
larly, a difference of stress is surely a difference of meaning. The semantic
difference between "Was it very big?" and "Was it very big?" 20 is clear.
Whether stress resides in intonation (where it is obviously part of meaning)
or in any other linguistic feature, the result is the same.
Rather than go on to discuss more examples-a procedure which can
never rule out the possibility of further ones which may be intractable-I
can perhaps here offer a dilemma: either the use of two expressions is
indistinguishable, in which case they are genuinely freely variant in every
sense, or their usage is distinguishable, in which case their meaning is dif-
ferent. The concept of free variation can, of course, never be more than
a hypothesis: we cannot rule out the possibility of hitting on a distinguish-
ing feature, however slight, in the usage of two expressions, even though
they may now seem in all contexts to be interchangeable. 2 1
So far I have argued that what are termed in ordinary language stylis-
tic variations are assimilable to differences of meaning and that, there-
fore, style is meaning. It remains to be shown that a dichotomy of two
distinguishable types of meaning cannot replace the dichotomy of style and
meaning. This is in any case a necessary adjunct to my argument so far, in
that even a strict dualist could argue that his use of meaning was dif-
ferent from my own, that I had merely shown that it was possible to use the
word meaning to refer to both sides of the dichotomy, and that no change
of substance was necessary for his theory if a terminology of two aspects
of meaning replaced the terminology of meaning and emphasis.
Koch advances a view of this kind, and he terms style split semantics. He
contrasts two words which differ stylistically, and he concludes that they
have one common area of meaning and one which is not common, a
semantic differential. His diagrammatic representation of this is two
oblongs, each representing one of the words: a large unshaded portion of
each represents the common meaning; a small shaded area, the semantic
differential. Yet for two distinguishable types of meaning to be involved,
the split in the semantics of a word would have to be made in the same way
in all its uses, which is clearly not the case. The semantic differential be-
tween two words is the result of the particular opposition of these two
20 The first merely asks for information concerning size; the second grants that the
object is big but asks for more precise information. These are thus two different ques-
tions, not the same question with a different stress.
21 I doubt the existence of completely freely varying expressions; this doubt, however,
has no bearing on my argument.
72 J. M. ELLIS

words, 22 and the same words contrasted to others would show different
areas of similar and dissimilar meaning. The separation of two elements of
the meaning of an expression achieved here is therefore not a typological
distinction but only one produced by an arbitrary juxtaposition of one
term and another. If, then, all stylistic variations are assimilable to
changes of meaning and if no typologically distinct type of meaning is
found, then style cannot be a useful concept. If this is so, some considera-
tion must be given to the fact that the concept of style does persist, often
seems useful, and is defined and redefined endlessly. The explanation of
this fact must be given in two parts, the first in relation to popular, the
second, to scholarly usage.
The ordinary language use cf the concept of style is dominated by pur-
pose in a given context. We can indeed say in ordinary language that there
are a dozen ways of saying the same thing, but this is because we wish to
achieve a quite delimitable purpose and can rule out all else as irrelevant.
If I am in command of a squad of soldiers and wish them to put some bul-
lets into a particular target, then it makes no difference for this purpose
whether I say "Commence firing," "Start firing," or "Begin firing." But the
particular phrase I use does other things than merely serve that purpose;
the whole situation has changed, slightly but noticeably. Similarly, the
phrases you can go now and get out can be ways of saying the same
thing; that is, they can, in part, achieve the same limited purpose. That
more and different things have been achieved, with the second particularly,
will become apparent on the next meeting of speaker and addressee. In
ordinary language, then, given a vast number of expressions in a language-
all with their own unique set of rules of use, but all with areas of overlap
with each other-I can in a given situation choose from a number to achieve
my purpose, and I may think of this operation as choosing between ways of
saying the same thing. Yet these expressions only exist as alternatives be-
cause of the unique context of my purpose, not because of their identity of
meaning, and, further, because I am able to write off the difference in mean-
ing between them as style (i.e., irrelevant for the moment). 23 Style is in
fact in ordinary language a way of avoiding those areas of the meaning of a
22 This point is developed below in my argument on the ordinary language function
of the concept of style.
23 Though the dualistic model has a useful function in ordinary discourse, it can of
course be dangerous here too: to think of a choice between expressions as a choice
among closely related meanings, rather than stylistic alternatives meaning the same
thing (ways of saying the same thing), would make a speaker more aware that he was do-
ing other things than merely achieving his present purpose-and these other things may
turn out to be important for other purposes which may be of interest to him outside his
present context.
 LINGUISTICS, LITERATURE, AND THE CONCEPT OF STYLE 73

word which are not considered essential to the speaker's dominant purpose.
These areas may be different for the same word in different contexts-a
further important reason why the theory of split semantics will not work.
This account of the popular usage of the concept of style shows its
inappropriateness as a concept in the theory of language. Several expres-
sions may share a function which is the dominant one for a speaker in a
given situation. Yet the linguist cannot for that reason ignore the difference
in function between the expressions, even in giving an account of their use
in a limited context. A speaker in saying "Commence firing" may not be
worried about what it is that has been achieved apart from producing fire;
but the student of language must be, and he cannot avoid noting the speak-
er's classification and in a sense alteration of the situation and his giving
information about himself.24
The ordinary language use of the concept of style, therefore, relies on the
existence of a crude purpose for a particular speech act which in turn
allows a simplification of the meaning of the expression used. It is immedi-
ately noticeable that this condition for the use of the concept in ordinary
discourse cannot occur in literary texts. Literature makes maximal use of
the patterns of meaning of expressions, instead of simplifying and sub-
ordinating them to a broad purpose. These texts are contemplated until
their details are absorbed and become important, while it is such details
that are made irrelevant in ordinary language by dualistic concepts such as
style. Literary texts are not limited speech acts with a crude purpose; they
are notoriously complex in their verbal structure and are not able to have
ascribed to them any simple statement of their purpose. Choosing a precise
meaning between close alternatives will be crucially important in these
texts much more often than in ordinary language. The concept of style
then becomes more and more inappropriate as we move from ordinary
discourse to the study of the language and finally to the study of literature;
it is precisely its limited appropriateness in ordinary discourse which
demonstrates its complete inappropriateness in literary study.
From this viewpoint, the persistence of the concept of style, and especi-
ally its elaboration and cultivation, might seem difficult to account for.
The mere retention is explained by the point made earlier: the kinds of
effect referred to as style are important for literary study, and to reject
the concept style may have seemed like rejecting these effects. It is cer-
tainly paradoxical that, as a result, a concept which originates in ordinary
language as a means of dismissing certain elements in the meaning of
an expression in order that others may be concentrated on should then
24 That this last type of information is that most commonly written off by a dualistic
model of essentials/extras is seen in the phrase "le style c'est l'homme."
74 J. M. ELLIS

be preserved precisely because those elements have restored to them the im-
portance which the concept of style denied them. The elaboration of the
definition of the concept follows from this contradictory position. A primi-
tive term from ordinary language has been accepted into the conceptual
vocabulary of a study while the theory which it implies is rejected. In the
puzzlement which results, the source of the trouble is attributed to the
complexity of the concept, so that to solve the problem, its meaning is
elaborated. The "real" meaning of an elusive and obscure word is then the
subject of the investigation, 2 5 and the primitive term has by this time had its
appearance quite changed by sophisticated confusion. However, the term
itself can never by this process be made anything other than a theoretically
primitive one. The general confusion over the definition of style is fur-
ther aggravated by a confusion between those sentences beginning "Style
is ... "which are properly definitions (i.e., attempts to give the meaning of
the term) and those which are statements of fact. "Style is a way of saying
something" is a definition; "Style ... is ... a product of structure" 26 is a
statement taking for granted the meaning of style (in one sense what style is)
and attempting to give us information about its basis, to tell us facts about
it. Yet in Enkvists's essay "On Defining Style" we read that a brief defini-
tion of style (italics mine) is: "The style of a text is the aggregate of the con-
textual probabilities of its linguistic items." 27 This confusion of radically
different functions of is 28 allows the already confused problem of lin-
guistic usage (the investigation of the meaning of the term) to be infinitely
25 Cf. the discussion by David G. Mowatt and by me of some concepts in Emil Staiger's
poetics: "Language, Metaphysics and Staiger's Critical Categories," Seminar, I (1965).
26 I take this statement from D. G. Mowatt and P. F. Dembowski, "Literary
Study and Linguistics," Canadian Journal of Linguistics, XI (1965), 44.
27 P. 28. Only on the basis of this methodological mistake could Enkvist make the
further statement: "Thus stylistic selection should not be defined as 'the choice between
items that mean more or less the same'" (p. 54). He is wrongly treating as alternatives a
statement about style and a definition of style and assuming that to accept the former
means rejecting the latter. His long treatment of the many "definitions" of style is viti-
ated by this mistake; but so are the "definitions" which he reports.
zs The two functions are shown most simply and clearly in the following pair of sen-
tences: "A ball is round" and "A ball is red." The first explains the meaning of the term
ball; the second could be used by someone explaining a game and has nothing to do
with the definition of the term ball. Still further complications are raised by the con-
fusion of stipulative definitions (terms introduced into an investigation with a stipulated
meaning and no other) and reportative definitions (reports of normal usage of a term
already in language). Thus, Sayee ("The Definition of the Term 'Style'") begins by saying
that words have the meaning their user chooses (true of stipulative, not of reportative,
definitions) and then goes on to consider style, a term which is already used popularly
and in scholarship. His definition, 'the use of language in a work of art' (p. 157), is only
interesting because it is an attempt to do something with a concept already established.
 LINGUISTICS, LITERATURE, AND TilE CONCEPT OF STYLE 75

more confused by bringing into the definition the results of research into
the designated phenomena.
If my argument so far has been correct, we should not speak of the
stylistic function of a linguistic item but, instead, of its precise function;
nor of the style of a piece of language, but instead of its precise meaning
and effect. I want now to conclude this article by considering how this
result affects the theory of the linguistic analysis of literature and its
relation to general linguistics.
The analysis of the precise function of a linguistic item or structure in a
text is in effect the analysis of its relation to all the other elements of the
text of which it is part. This is a study, then, of the combination oflinguis-
tic structures in a text. That combination, and not isolation, is the focus of
attention in the linguistic study of literature is a very important, but fre-
quently ignored, fact. The point of this distinction, however, is made clearer
if it is seen that it relates to distinct but necessarily complementary activ-
ities within linguistic study: on the one hand, the investigation of a
language, and on the other hand, the investigation of a text.29
It is proper, it seems to me, to consider that the study of language in-
cludes both the general description of the patterns and possibilities of a
language and also the description of particular texts or exploitations of the
patterns and possibilities of that language. The first is an analytic exercise
which isolates the patterns of the language, and it produces grammars and
dictionaries. The second is concerned with the synthesis of linguistic
structures in one text. Both exercises involve analysis; in the first, however,
analysis is an end in itself, while in the second, the end is the investigation
of the combination and interaction of structures found by analysis. The
first might be termed analytic linguistics; the second, synthetic linguis-
tics. Both are concerned with meaning and structure: the first with the
inventory of structures within a given language and with their functions in
general; the second with the interaction of a given set of meaningful struc-
tures and with the resulting structure and function of the whole text. Put
another way, the first generalizes about the possibilities of the language
and classifies them; the second particularizes about the relation in a text of
a given set of them.

and already meaning something important for literary study. If taken seriously, this
stipulation abolishes the concept of style, though it proposes to go on using the word in
an almost empty sense-an unnecessarily confusing procedure; that the word is used is
evidence that the ordinary language meaning is still being exploited, however.
29 This part of my argument owes something to Saussure's related distinction be-
tween langue and parole; in the particular use made of the distinction here, though, I
owe rather more to conversations over many years with D. G. Mowatt.
76 J. M. ELLIS

An analogy can be drawn between the relation of these two aspects of

linguistics and that of physics and engineering. A physicist attempts to
isolate the mechanisms of the physical world. An engineer, instead of
analyzing the basic patterns of the physical world, is concerned with the
functioning together of these isolated patterns in a unique structure. He is
concerned with its equilibrium, with how each element contributes to the
stability of the structure, and with how general physical laws interact in the
finished structure. Similarly, critics of literature are concerned with the
interaction of general features of a language to form a unique structure.
The study of literature is only part of synthetic linguistics, which covers
the investigation of any text; though literature, as a body of texts thought
of as being especially interesting combinations of the structures of a
language, understandably looms large as material for consideration. There
are, however, many other types of text which will be worth study: political
speeches, scientific writing, and so on. All will involve combinative study.
Iflinguistics is the study oflanguage, then it must include both the study
of the elementary and the complex structures oflanguage, both the isolation
and the combination of its elements. There is no reason to suppose that the
study should not include generalizing and particularizing, even though the
overwhelming majority of general linguists pursue exclusively the general
descriptive analysis of a language. The study ofliterature, then, is a part of
one aspect of linguistics. 30
Yet, since nearly all work in general linguistics is (in my terminology)
analyticlinguistics, it has seemed easy to equate the two, and, if this is done,
linguistics (=analytic linguistics) is indeed not literary criticism. Critics
and linguists alike have generally agreed, on this unreal basis, to an unreal
separation between them, missing the fact that each activity entails the
other. To study the combination of linguistic structures and their functions
in a text, for example, presupposes their identification and classification
in general terms.
The most unfortunate aspect of the identification of linguistics with
analytic linguistics, however, is the tendency to see any distinctively
linguistic study of literature on the analytic model :3 1 isolative, classifica-
3o I thus agree with Jakobson ("Closing Statement," p. 350): "Since linguistics is
the global science of verbal structure, poetics may be regarded as an integral part of
linguistics." Furthermore, my distinction of two types of linguistic study is not entirely
inconsistent with his dictum (p. 358): "The poetic function projects tlie principle of
equivalence from the axis of selection into the axis of combination."
31 The closing section of my argument is of necessity only a brief and schematic
account of the implications of my view of style for the relation of linguistics to literary
study. A fuller treatment of this relation is reserved for a later date. What is offered here
is an indication of a next stage in the argument, rather than a demonstration of it.
 LINGUISTICS, LITERATURE, AND THE CONCEPT OF STYLE 77

tory, and exemplificatory. The deviation theory which I have discussed

above, for example, has through its long and varied career posed questions
which are irrelevant, because they are the questions of analytic linguistics,
where analysis is the end instead of a stage in the investigation leading to
combinative statements. Thus, attempts are made to isolate the distinctive
features of style, to describe it as deviant from the grammar, 32 or to treat it
as a single device, all of which is very much beside the point.
Another all too popular view of the contribution of linguistics to litera-
ture which rests on this limited notion of linguistics is the view that
linguistic analysis of literature is a precritical activity: it provides data for
the critic to use as his critical intelligence sees fit.33 Against this view it
only needs to be said that isolation of linguistic features of a literary text is
either a pointless exercise in analytic linguistics (a gratuitous exemplifica-
tion of the grammar of the language) or it is a strangely incomplete job of
synthetic linguistics. It is, on the one hand, as absurd to suppose that an
investigator of the combination of structures in a text should not be able
to isolate and describe them himself (for how else could he deal with them
properly?) as it is, on the other hand, to suppose that those elements of a
text which are important for its total structure can be isolated without a
view of what that total structure is. This theory, then, arises because
linguistics is illegitimately restricted to the isolative activity which is only
a part of the study of language. 34

32 Riffaterre goes even further and separates style and language: we must avoid, he
says, "the confusion between style and language" ("Criteria," p. 154). It is, of course,
only necessary to insist that style is a matter of verbal structure. Riffaterre is misled
partly by his rigid dualism (language is the "vehicle" of style) but also by the absence of a
distinction between analytic and synthetic linguistics: "stylistic facts ... must have a
specific character, since otherwise they could not be distinguished from linguistic facts"
(ibid.). Here linguistic facts is obviously thought of as equivalent to isolated linguistic
facts, but the logic of the argument breaks down when we substitute this expanded
phrase. The specific character sought is the combination of isolated linguistic facts; that
is, it is still linguistic.
33 This is too common a theory for it to be attributed particularly to any one author, but
it seems especially popular among British linguists. Its most recent appearance seems to
be in Fowler's Essays on Style and Language (seen. 2). In a review (Journal ofLinguistics,
IV [1968], 109-115), Donald C. Freeman welcomes the book's taking this view and gives
a short summary of it. Freeman's view is that a "critical intelligence" must inform the
linguist's isolations. Yet the separation of roles has now collapsed: someone with a
critical intelligence would seem to be indistinguishable from a critic.
34 I assume that there is no need in this context to discuss the view common among
critics that analysis of the verbal structure of a text is not a study of its meaning, with its
corollary that linguistics can only be concerned with the mechanisms, but not the mes-
sage, of a piece of literature.
78 J. M. ELLIS

Signs of a move from an isolative to a combinative view of the linguistic

analysis of literary texts can be seen: a number of recent works have set up
fairly simple combinative models, and terms like cohesion 35 and con-
gruence have begun to appear. But in general, the concept of style, help-
ing to support a view of linguistics which is solely isolative, still inhibits
linguistic analysis of literature. Linguists have not aroused admiration
among critics for their contributions to literary study recently; much pre-
cise and technical analysis seems to have contributed little that is useful. If
this is so, it is because the concept of style has been the foundation on
which linguists have built. We shall no doubt continue to use the concept
of style in ordinary discourse; however, its use as a basis for linguistic and
literary study should, if I am correct, cease.

Adlai E. Stevenson College

University of California
Santa Cruz, California 95060

35 The term is M. A. K. Halliday's ("The Linguistic Study of Literary Texts," Pro-

ceedings of the Ninth International Congress of Linguists [Cambridge, Mass., 1962],
p. 303): "While insisting that stylistic studies use the same methods and categories as
non-literary descriptions, we must make the proviso that such studies may require new
alignments or groupings of descriptive categories, through which the special properties
of a text may be recognized. This may include the bringing together of categories and
items described at different levels as well as those scattered throughout the description of
any one level. An example of such a grouping, in which various grammatical and lexical
features are brought together, is 'cohesion'." The study of the combination of linguistic
structures at various levels is, therefore, formal and linguistic. I should also propose that
stylistic need not be opposed to nonliterary and that cohesion be regarded in a more general
way: not as a limited abstraction from particular texts, but as the general framework for
the investigation of any text; in the investigation of a text, the focus is on the cohesion of
all of its elements.